Seeking the optimal accuracy-rate equilibrium in face recognition

Considerable facial data is often compressed prior to analysis to accommodate limitations in transmission or storage capacities. However, this compression may lead to the loss of crucial identity details, thereby diminishing the effectiveness of facial recognition (FR) systems. In this study, we aim to establish an optimal accuracy-rate equilibrium (ARE) to maximize the compression ratio without substantially compromising the performance of the FR system. We first investigate the effect of image compression in deep FR. Based on the definition of ARE in FR, we proposed a method to locate the ARE values of face images with true acceptance rate and false accept rate. Subsequently, we develop an ARE prediction method for the FR system (ARE-FR), which automatically infers ARE images of face images. The goal of the proposed ARE-FR is to maximize redundancy removal without impairment of robust identity information. Considering that high-level semantic features effectively capture crucial identity information, we force the proposed ARE-FR to focus only on the features in identity-related regions when predicting the ARE images. These features are derived from the interactive relationships between deep and shallow features. Experimental results have demonstrated that combining our proposed ARE-FR with the image coding algorithm is capable of saving more bits while maintaining the performance of the FR system.